基于图神经网络的天地一体化网络建模及性能预测  

Modeling and Performance Prediction of Space-integratedGround Network Based on Graph Neural Network

作　　者：潘成胜 沈凌宇 赵晨 崔骁松 PAN Chengsheng;SHEN Lingyu;ZHAO Chen;CUI Xiaosong(School of Electronics and Information Engineering,Nanjing University of Information Science and Technology,Nanjing 210044,China)

机构地区：[1]南京信息工程大学电子与信息工程学院,南京210044

出　　处：《火力与指挥控制》2025年第2期13-20,共8页Fire Control & Command Control

基　　金：国家自然科学基金资助项目(61931004)。

摘　　要：随着新型作战装备的不断涌现,我军指挥控制网络呈天地一体化趋向,作战元素增多对指挥控制网络业务低时延、低抖动的传输能力提出了更高要求。为解决异构一体化指挥控制网络难以对流量的复杂特性形成准确约束以及网络建模困难的问题,提出一种基于图神经网络和注意力机制融合的网络性能预测模型,以实现对天地一体化指挥控制网络中信息传输时延和抖动性能的精准预测。实验表明,针对指挥控制信息传输性能,该模型具有良好的预测效果。With the continuous emergence of new combat forces,the land air defense command and control network has shown a trend of space-integrated-ground,and the increase in combat elements has put forward higher requirements for the low-delay and low-jitter transmission capabilities of command and control network services.To address the difficulties of accurately constraining the complex characteristics of traffic and modeling in heterogeneously integrated network,a network performance prediction model based on the fusion of graph neural networks and attention mechanisms is proposed to achieve accurate prediction of traffic transmission delay and jitter performance in the integrated air defense command and control network.Experiments have shown that the model has good predictive performance for air defense ground combat command and control traffic.

关 键 词：网络性能预测 天地一体化 图神经网络 深度学习 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]